Thermodynamic lake ice models are valuable tools in the simulation of ice formation, growth, and decay. Appropriate application of these models necessitates a thorough understanding of model physics. Here, we examine the physics of two thermodynamic lake ice models, the Canadian lake ice model (CLIMo) and the High-Resolution Snow and Ice Model (HIGHTSI), for understanding key drivers and limitations in modelling of ice evolution. A cold bias in modelled surface temperatures was found to control differences in ice evolution through differences in the magnitudes of radiative and turbulent fluxes. Simplified snow physics and precipitation forcings were found to control the simulation of snow-ice. Break-up dates were highly sensitive to the selected melt albedo scheme. Freeze-up dates were controlled by model specific calibration or initialization procedures. Recommendations for advancements to current model processes are presented for future developments to thermodynamic ice models.